Since handling of articles involves heavy labor, in recent years, it is becoming difficult to secure human resources. Therefore, there is a demand for automation of unloading operations, such as depalletizing. In a conventional article handling device for performing unloading operations, an optical sensor, such as an image sensor which is capable of acquiring a wide range of information at once, has been used to recognize articles to be unloaded. However, in cases such as articles having complicated patterns, articles arranged closely in contact with each other without a gap, or the edges of articles being hidden by obstacles such as ropes and labels, due to the nature of optical sensors, recognition would not be performed correctly. If the articles are gripped based on an incorrect recognition result, the gripping may fail.